Navigating AI and ML: A Roadmap for Engineers and other STEM Professionals

On the opening day of the 2024 Remarkable Conference by the Vector Institute for Artificial Intelligence (AI), a timely question was posed to Cody Coleman following his keynote presentation: How an engineer new to Artificial Intelligence and Machine Learning (ML) should start exploring the overwhelming abundance of online resources? While Coleman offered insightful avenues for engagement, this article provides a roadmap to navigate the initial stages of the AI & ML journey.

Photo by Clemens van Lay on Unsplash

Embarking on this journey offers a transformative pathway for engineers, even those just starting with coding. For those who are still in school, I highly recommend taking some coding, AI and ML courses offered by your institution. Professional engineers and peers in science, technology, and mathematics (STEM) disciplines considering a more formal educational path, pursuing a master’s degree in these fields can significantly strengthen your expertise and career prospects. Many universities worldwide now offer (in-person and on-line) specialized programs focused on AI and ML, catering to the growing demands for skilled professionals in these areas. At my university — Ontario Tech, we offer many graduate-level Engineering programs that provide students the opportunity to take courses in programming and machine learning; and many of our undergraduate Engineering programs include courses or a specialization in AI engineering. Pursuing a master’s degree provides an opportunity for deeper specialization and research under the guidance of experts in the field, paving the way for innovation and advanced learning.

The rest of this article provides a self-learning pathway, guiding you through a curated selection of resources and strategies designed to empower your journey into AI and ML.

Assess Your Math & Stats Background

Engineers and their counterparts in the broader STEM disciplines have a unique edge, thanks to their robust mathematical foundation in areas like Linear Algebra, Calculus, and Probability & Statistics, which are the foundations for AI and ML. This foundational knowledge not only facilitates a smoother transition into the domain but also empowers you to excel in developing complex models and understanding algorithmic intricacies. If you do not have the math skills, you can brush up on these fundamentals using Khan Academy, or this freely available book Mathematics for Machine Learning by Marc Peter Deisenroth, A. Aldo Faisal, and Cheng Soon Ong — it is a comprehensive resource that covers the necessary mathematical skills for machine learning.

Do you know Python? The Lingua Franca of ML

Most of the ML frameworks such as scikit-learn, TensorFlow, Keras, and PyTorch are largely implemented in or supported by the Python programming language. Python offers a rich ecosystem of libraries and packages designed to facilitate machine learning and data analysis. If you have not used Python before, a good place to start is Automate the Boring Stuff with Python, which offers a practical approach to learning Python. If you know the basics of Python then you should learn about three prominent Python packages for working with large data files and data analysis:

• NumPy: Fundamental package for scientific computing with Python, providing support for large, multi-dimensional arrays and matrices, along with a collection of mathematical functions to operate on these arrays.
• Pandas: Provides high-performance, easy-to-use data structures, and data analysis tools. It’s particularly suited for data manipulation and analysis.
• Matplotlib: A plotting library for creating static, interactive, and animated visualizations in Python.

A good place to start with those Python packages is O’Reilly’s Python Data Science Handbook by Jake VanderPlas.

For those who prefer video-based learning, I recommend exploring Python Data Science — A Free 12-Hour Course for Beginners offered by freeCodeCamp.

Learn the Fundamentals of AI and ML

To learn the fundamentals of AI and ML, I recommend the following three steps:

1. The book “Artificial Intelligence: A Modern Approach” by Stuart Russell and Peter Norvig, is a classic text that is widely regarded as one of the most comprehensive textbooks for understanding the fundamental theories and concepts of AI; we use this book in our course “Introduction to Artificial Intelligence” for third year students in software engineering. While it’s highly recommended for its depth and breadth of coverage, it’s not strictly mandatory for everyone wanting to get into AI and ML, so those pressed for time or seeking more direct engagement with machine learning concepts can proceed to the next step.
2. For a comprehensive foundation on ML, the Machine Learning course by Andrew Ng on Coursera is highly recommended. It covers key concepts and algorithms.
3. For a hands-on approach, consider “Hands-On Machine Learning with Scikit-Learn, Keras & TensorFlow” by Aurélien Géron, which guides you through building ML models with popular libraries.

Despite the many online resources available, reading “Pattern Recognition and Machine Learning” by Christopher Bishop is still worthwhile due to its comprehensive coverage of fundamental machine learning concepts, which are essential for understanding how deep learning works. Although the book was published before the recent explosion in deep learning, its focus on the principles of machine learning makes it a valuable resource covering a broad spectrum of topics within the field, including Bayesian networks, decision trees, regression, and neural networks.

Dive into Deep Learning

After grasping the basics of ML, explore deep learning through the freely available MIT Press “Deep Learning” book by Ian Goodfellow and Yoshua Bengio and Aaron Courville.

If you are a visual learner, an alternative would be the “Deep Learning Specialization” by Andrew Ng on Coursera. This series of courses will introduce you to the workings of neural networks and how they’re applied.

Choose your Application Domain

Artificial intelligence can be applied to many application domains, each with its own unique challenges and innovations. Some of the application domains include:

• Computer Vision: Focuses on enabling machines to interpret and make decisions based on visual data.
• Natural Language Processing (NLP): Deals with the interaction between computers and humans using natural language.
• Robotics: Integrates AI to develop robots that can perform tasks autonomously.
• Autonomous Vehicles: Involves AI in the development of self-driving cars and drones.

You should identify the area(s) that internet you and focus on them to build specialized ML expertise in that application domain through hands-on practice.

Hands-on Practice

Applying your knowledge through projects is the best way to learn, here are some recommendations for doing so:

• Kaggle competitions: Kaggle offers a plethora of datasets and beginner competitions that provide structured datasets and evaluation metrics, allowing you to apply your learnings to solve practical problems and compete with others. Don’t be intimidated by the leaderboard — focus on learning and progress.
• Personal side projects: As you gain confidence, embark on personal projects that challenge you and align with your interests. Perhaps you build a recommendation engine for movies, books, or music; or an application to analyze sports data to predict winning teams. The possibilities are endless!
• Open-source contributions: Give back to the community by contributing to beginner-friendly open-source projects on GitHub related to your chosen AI subfield. This allows you to learn from experienced developers, hone your coding skills, and gain valuable exposure to real-world AI projects.

Networking on the Network

Stay updated with prominent publications like The Gradient and Towards Data Science and join AI and ML communities on platforms like LinkedIn, Reddit, and GitHub. Participating in forums and discussions can provide insights and help you stay updated with the latest trends and technologies.

As you embark on your machine learning journey, remember: (1) dedicate regular time to learning and practicing; even modest but consistent efforts can accumulate into significant advancement; and (2) don’t be afraid to fail, it is a vital part of the learning process.

To probe further

• Teachable Machine: A web-based tool to demonstrate AI & ML, encouraging experimentation and innovation — no coding required.
• TensorFlow Playground: Visually explore machine learning concepts and build simple models directly in your browser.
• Hugging Face: Open-source library for natural language processing with pre-trained models and interactive demos.
• AI in Business: Learn how AI is transforming various industries and businesses with practical insights and interviews.
• Machine Learning subreddit: Join the online community of learners and experts on Reddit for discussions, tips, and resources.